[cairo] intersect_lines in Bentley Ottman implementation

[Vladimir Vukicevic]

On Aug 4, 2008, at 12:14 AM, Antoine Azar wrote:
> The floating point optim gave an average of 11%. This is of course  
> dependant
> on how many edge intersections are present in the test (about none  
> in the
> performance suite). We would see a much larger performance gain if  
> there
> were more edge intersections to compute, but as Soeren pointed out,  
> those
> are probably rare cases.

Those are probably rare, but those are also the ones where overall  
performance may be significantly slower than the common/simple cases..  
and a large speedup may be more noticeable.

     - Vlad